Dear Chalmers
I have come across a similar question like "Jin and Wang (2014) via createItem()" [
https://groups.google.com/g/mirt-package/c/JND5k0cQRNA/m/mc7ychJSAQAJ], can you help with it? I would like to include
w as a constant in the likelihood of a single participant. Specifically: for item parameter estimation by EM or similar algorithms, we need to input the initial value of item parameter:
x_0, at this time, make
w=1 for all
participants, that is, no change in marginal likelihood, after that, we will estimate the item parameter under the first iteration:
x_1 =
x_0 - grad(
x_0)/hessian(
x_0), and I will use the new item parameter
x_1 and the response data of each
participant to calculate the corresponding personfit statistic, and at the next iteration of the item parameters, I will use each
participant's personfit statistic to update
w (which is also a constant ), then the marginal likelihood will change (which in the EM algorithm,
w only affects the value of the expected number), and then based on the new item parameter
x_2 for the next iteration, calculate the personfit statistic to update
w again. Until the criterion is satisfied, and output the item parameters to be estimated.
how can I implement such a parameter estimation process ? I really need your help to implement this process in mirt. Looking forward to your reply.
Sincerely
Liu Kang
